Inter-American Development Bank
Banco Interamericano de Desarrollo (BID)
Research Department
Departamento de Investigación
Working Paper #684
RG-N3338
Education and Democratic Preferences
by
Alberto Chong*
Mark Gradstein**
* Inter-American Development Bank
** Ben Gurion University
June 2009
Cataloging-in-Publication data provided by the
Inter-American Development Bank
Felipe Herrera Library
Chong, Alberto.
Education and democratic preferences / by Alberto Chong, Mark Gradstein.
p. cm. (Research Department Working Papers ; 684)
Includes bibliographical references.
1. Democracy and education. I. Gradstein, Mark. II. Inter-American Development Bank.
Research Dept. IV. Title. V. Series.
LC71 .C466 2009
370.115 C466----dc22
©2009
Inter-American Development Bank
1300 New York Avenue, N.W.
Washington, DC 20577
The views and interpretations in this document are those of the authors and should not be
attributed to the Inter-American Development Bank, or to any individual acting on its behalf.
This paper may be freely reproduced provided credit is given to the Research Department, InterAmerican Development Bank.
The Research Department (RES) produces a quarterly newsletter, IDEA (Ideas for Development
in the Americas), as well as working papers and books on diverse economic issues. To obtain a
complete list of RES publications, and read or download them please visit our web site at:
http://www.iadb.org/res.
2
Abstract*
This paper examines the causal link between education and democracy.
Motivated by a model whereby educated individuals are in a better position to
assess the effects of public policies and hence favor democracy where their
opinions matter, the empirical analysis uses World Values Surveys to study the
link between education and democratic attitudes. Controlling for a variety of
characteristics, the paper finds that higher education levels tend to result in prodemocracy views. These results hold across countries with different levels of
democracy, thus rejecting the hypothesis that indoctrination through education is
an effective tool in non-democratic countries.
Keywords: Education, democracy
JEL Classifications: I20, I30, Y80
*
Chong: Inter-American Development Bank. Email: [email protected]; Gradstein: Ben Gurion University, Israel;
CEPR; CESifo; IZA. Email: [email protected]; Vanessa Rios provided very able research assistance.
3
1. Introduction
Education, by fostering civic attitudes and cultivating rational informed opinions, has been
commonly believed to promote pro-democracy views (Mann, 1846; Dewey, 1916). The more
recent stipulation by Lipset (1959) that education is a prerequisite of a democratic society was
echoed by Milton Friedman, “A stable and democratic society is impossible without a minimum
degree of literacy and knowledge on the part of most citizens and without widespread acceptance
of some common set of values. Education can contribute to both” (Friedman, 1962, 86). This
link is often assumed in theoretical models of economic development as well. For example,
Bourguignon and Verdier, 2000, stipulate that education tends to make people more critical of
the prevailing autocratic regime, thus implying that investment in education by such a regime
may ultimately undermine it.
Empirical work, mostly in a cross-country context, similarly documents a positive
association between education and democratic institutions (Barro, 1999; Glaeser et al., 2004;
Glaeser, Ponzetto and Shleifer, 2007). This literature, however, has difficulties addressing
endogeneity and reverse causality. Also, there is evidence that education is positively related to
measures of civic participation (Campante and Chor, 2008; Dee, 2004; Glaeser, Ponzetto and
Shleifer, 2007; Milligan, Moretti and Oreopoulos, 2004).1 A dissenting view is presented in
Acemoglu et al. (2005), where, when controlling for fixed effects, education turns out to be
unrelated to democracy; see, however, Bobba and Coviello (2007), for qualifications in regard to
the econometric technique used.
Another important observation, in Lott (1999) is that totalitarian regimes tend to make
substantial investments in education, presumably as the means of indoctrination. A theory on
which this is based, more fully exposed in Lott (1990) provides a link between indoctrination
through the provision of public schooling and regime legitimacy. Spilimbergo (2008) finds that
education acquired in foreign countries promotes democracy domestically, provided that the
foreign country is democratic; when the foreign country is non-democratic, no effect is detected.
In this paper, we make use of international survey data to revisit the relationship between
education and pro-democracy opinions. To shape ideas, consider the following two views that
emerge from the literature discussed above.
One view emphasizes indoctrination, seeing
1
An issue here is that civic participation cannot be readily interpreted as being necessarily pro-democratic—think of
neo-Nazi rallies, for example.
4
education as a means of promoting attitudes. Then the ultimate effect on democratic disposition
depends on the values instilled, which may well differ between democracies and nondemocracies: while the former may promote pro-democratic views, the latter may indoctrinate
against such views. An alternative theory, formally developed in the paper, is that education
reduces the cost of acquiring an informed opinion, needed to assess public policies. In this case,
education’s effect on pro-democracy views is positive independently of the regime type, as
educated individuals have a comparative advantage under democracy in being able to affect
policies.2 Thus, while both theories generate a positive effect of education on pro-democracy
attitudes under democracies, their predictions on this relationship in non-democratic regimes
differ.
We contribute to the understanding of these issues by first formalizing the idea, often put
forward by educators, that education delivers informational benefits relevant for democracy. To
this end, we provide an analytical framework where better judgmental capacity, assumed to be
correlated with education, leads to better policy choices.
Consequently, better-educated
individuals have a comparative advantage in a democracy, where popular voice plays a role. In
the model, this mechanism ensures that education leads to pro-democracy attitudes.
We then examine the World Values Surveys data, which contain detailed questions about
democratic preferences. We find that education is correlated with these preferences even after
controlling for a variety of personal characteristics. Although reverse causality is much less of
an issue here than in cross-country settings, and despite employing a relatively rich set of
personal characteristics as controls, we also use instrumental variables to primarily address
possibly omitted variables.
The results reaffirm the positive effect of education on pro-
democratic attitudes. Further, we find that this relationship holds across both democracies and
non-democracies. In other words, a country’s level of democracy, when interacted with the level
of schooling, turns out to be statistically insignificant, contrary to the theory of Lott (1990) that
educated individuals in non-democratic countries, being subject to intense indoctrination, may be
less supportive of democracy than less educated individuals.
The paper proceeds as follows. The next section presents our analytical framework. The
data and empirical strategy are discussed in Section 3. Our main empirical results are contained
in Section 4, and Section 5 concludes.
2
This theory is consistent with the view presented in the opening paragraph above.
5
2. Analytical Framework
2.1. The Model
Consider an economy populated with a continuum of citizens of a unit measure, and a ruler; the
role of the latter is to implement public projects. The population of individual citizens forms a
legislature. The citizens are initially endowed with identical incomes normalized to zero.3 A
potent ruler expropriates the individuals’ income and provides a public good. We let e denote
the amount of expropriation and b the net benefit from the public good accrued to every citizen.4
The benefit is assumed to be ex ante unknown and distributed according to the distribution F(b)
and lies in the interval [0,∞).
The status quo refers to the situation where there is no public good and no expropriation,
in which b=e=0. We assume for analytical simplicity that the utilities are linear. The utilities of
the citizens and the ruler can then respectively be written as follows:
U = (-e + b)Δ
(1)
and
R=eΔ
(2)
where Δ=1 if the ruler is potent, and Δ=0 if she is blocked and the status quo prevails. Note that
under the first best outcome and without strategic distortions, the public good is provided, and
the aggregate surplus is b.
The level of democracy, denoted v, is the probability of holding the ruler accountable; 1-v
is the probability with which the ruler acts autonomously. In the latter case, the public project is
implemented and expropriation takes place. In the former case, the ruler submits a proposal on
whether to carry the project out and on the amount of expropriation. The project goes through if
approved by the legislature, but if it is blocked then the status quo prevails.
Maintaining a given accountability level is assumed costly for the citizens, and this cost
differs in relation to their education levels.
The idea is that more accountability requires
maintaining and processing of diverse sources of information as well as a high level of public
discourse, the costs of which are smaller for educated people.
3
We will discuss extensions to heterogeneous agents later.
This refers both to pure public goods and to private goods that for a variety of reasons are state provided, such as
education. We will refer to all these cases, slightly abusing the terminology, as public goods or public projects.
4
6
In particular, the accountability cost function is C(hi,v) = c(hi)v2/2, c’<0, 0<v<1, where hi
is individual i’s education, and the assumption on c ensures that better educated individuals incur
a lower cost for a given level of accountability.
In the first stage, individuals set the level of accountability v. Then, with the probability 1v the ruler’s proposal is implemented, whereas with the probability v it undergoes a review by
the legislature and can be either approved or blocked. We study the subgame perfect equilibria
of the resulting game.
2.2. Analysis
Proceeding backwards, we first examine when the ruler’s proposal is approved. This will be the
case when the citizens’ utility under the proposal exceeds the status quo level, or when -e + b >
0. It then follows that the realization of the value of the public project that makes the citizens
just indifferent between approving the proposal and rejecting it is b = e, and the resulting
probabilities of its rejection and approval, respectively, are F(e) and 1-F(e).
For a given level of accountability, the ruler’s expected utility is
∞
ER = (1-v)e + v ∫ edF (b) = e - veF(e)
e
and its maximization with respect to e yields the first order condition
1 – v(F + eF’) = 0
that determines the amount of expropriation proposed by the ruler; differentiation reveals that it
decreases in v.
We then write the citizens’ expected utility as follows:
∞
∞
0
e
EUi(v) = - c(hi) v /2 + (1-v) ∫ (b − e)dF (b) + v ∫ (b − e)dF (b)
2
(3)
Its differentiation yields the first order condition for the favored level of political accountability
by citizen i:5
5
Internal solutions are assumed throughout.
7
∞
∞
0
e
- c(hi)v - ∫ (b − e)dF (b) + ∫ (b − e)dF (b) - [1-v+v(1-F(e)]de/dv =
e
- c(hi)v + ∫ (e − b)dF (b) = - c(hi)v + F(e) - [1-vF(e)]de/dv = 0
(4)
0
and the second order condition is assumed to hold. Equation (4) implicitly determines v(hi), and
its total differentiation reveals that ∂v(hi)/∂hi > 0, implying that higher education implies
increased preference for political accountability.
Suppose now that a reduced form version of a political process is used to determine the
accountability level, whereby a weighted function of individual utilities, ∫ ω (hi )U i di , is
maximized. We then let hd denoted the education level of the decisive individual and the first
order condition is
- c(hd)v + F(e) - [1-vF(e)]de/dv = 0
and total differentiation yields that the more educated the decisive individual the higher is the
equilibrium level of accountability.
Summarizing,
Proposition 1. Education is positively associated with individual pro-democracy views and with
collective support for democracy.
2.3. Education, Policy Assessment, and Pro-Democracy Attitudes
Rulers often use deception, manipulating information to influence citizens’ attitudes in order, for
example, to affect their voting behavior. This phenomenon may exist in both democratic and
non-democratic settings, although Lott (1990, 1999) suggests that it may be particularly relevant
in the latter. To illustrate how this may work, suppose now that the population is divided into
educated and uneducated citizenry, and let μ denote the fraction of the former. The public good
is parameterized with q, its quality, so that the distribution of benefits, F(q,b), Fq >0, shifts
rightwards with an increase in q, which is assumed to be privately known to the ruler.
8
In the first stage, the ruler sends a message m(q) about the true value of q, bearing a cost
of deception that increases in the deviation from the true value, φ(m-q)2/2, φ>0.6 This cost
captures, in particular, social disapproval in case deception is revealed, as well as the resources
needed for excessive manipulation of information. The difference between the educated and the
uneducated individuals is that, while the former behave in a Bayesian manner, critically
evaluating the message and forming posterior beliefs about the true value, the latter individuals
are naïve and simply believe the message. To focus on this difference, we now assume that all
individuals share equally in the cost of maintaining accountability, c(hi) = 1. To further simplify,
we assume that the amount of the ruler’s expropriation, e, is given.7
We will employ the fact that the above game is a special case of the one studied in Kartik,
Ottaviania and Squintani (2007). As in that paper, we focus on separating perfect Bayesian Nash
equilibria. Adopted to this context, such an equilibrium is defined by educated individuals’
beliefs formed under Bayes’ rule and the ruler’s message that maximizes her utility given these
beliefs and correctly anticipating individuals’ decisions.
The analysis proceeds backwards. The approval rule of the ruler’s proposal by the citizens
is as before, and it is approved when b>e. We then write the individual utilities as follows:
∞
∞
0
e
EUi(v) = - v + (1-v) ∫ (b − e)dFi (Qi (m), b) + v ∫ (b − e)dF (Qi (m), b)
(5)
where Qi(m) are individual i’s beliefs and F(Qi (m),b) is individual i’s assessment of the
distribution of benefits under these beliefs. Differentiating, we obtain the first order condition
for the preferred level of accountability by individual i:
∞
∞
0
e
−v - ∫ (b − e)dF (Qi (m), b) + ∫ (b − e)dF (Qi (m), b) = 0
(6)
Recall that uneducated individuals naively believe the message m; Kartik, Ottaviania and
Squintani (2007) show that in equilibrium, Bayesian individuals correctly invert the message,
deducing the true value q. Thus, Qi(m(q))=m if i is uneducated, and Qi(m(q))=q if i is educated.
Further, differentiation of (6) reveals that the preferred level of accountability is a decreasing
function of the perceived quality of the public good, dv/dQi < 0.
6
7
This assumption is stronger than needed.
This does not affect any of the qualitative results.
9
Letting ω(μ) be the relative weight of the fraction of educated individuals, an increasing
function, we write the first order condition for the collective choice of accountability:
∞
∞
0
e
−v + ω(μ)[- ∫ (b − e)dF (q, b) + ∫ (b − e)dF (q, b) ] +
∞
∞
0
e
(1-ω(μ)) [- ∫ (b − e)dF (m, b) + ∫ (b − e)dF (m, b) ] = 0 (7)
and total differentiation reveals that dv/dm < 0 and that dv2/dmdμ > 0. Thus, accountability
decreases with the level of the message, but less so the greater is the fraction of educated
individuals.
We now employ the fact (see Theorem 1 in Kartik, Ottaviani and Squintani, 2007) that
there is a unique separating equilibrium, with m(q) strictly increasing; we then write the utility of
the ruler when issuing a message m whereas the true value is q as follows:
∞
ER(m,q) = -φ(m-q)2/2 + ω(μ) [(1-v(q))e + v(q) ∫ edF (q, b) ] +
e
∞
(1- ω(μ))[(1-v(m))e + v(m) ∫ edF (q, b) ] = -φ(m-q)2 + e - ω(μ) v(q)eF(q,e) e
(1- ω(μ))v(m)(1-F(q,e))
(8)
where ω(μ) is the relative weight of the fraction of educated individuals, which is an increasing
function.
The first order condition for the equilibrium message is
-φ(m-q) - (1- ω(μ))(dv(m)/dm)(1-F(q,b)) = 0
(9)
and the second order condition holds. Note that, since dv(m)/dm < 0, m>q, so that the ruler’s
message is always exaggerated. Further, totally differentiating (8) we obtain that dm/dμ < 0, so
that the extent of exaggeration is a declining function of the fraction of educated individuals.
Summarizing these results,
Proposition 2.
In equilibrium, the ruler’s message about the public good’s quality is
exaggerated, with the extent of exaggeration decreasing in the fraction of educated individuals.
Educated individuals favor a higher level of accountability than do uneducated ones, so that the
10
equilibrium level of accountability is an increasing function of the fraction of educated
individuals.
These results complement those of the previous sub-section in indicating that educated
individuals have a comparative advantage in making policy inferences under democracy, even
when subject to manipulation. Education, again, is seen to promote democratic preferences,
which also has implications for the collective choice, so that democracy is more likely to emerge
the more educated society is. In this sense, the model results are fully consistent with the
sentiments by prominent commentators cited above.
3. Data and Empirical Strategy
In order to test the empirical implications of the theoretical model, we use the following
benchmark specification:
Democracy ij = β Schooling ij + γX ij + λZ j + u ij
(10)
In this equation, Democracy is the dependent variable and represents the preference for
democracy by individual i in country j. According to the model presented above, we focus on
education of individuals as our main variable of interest (Schooling). We also include vector X
that represents additional individual-level explanatory variables, vector Z that represents country
and year dummies, and uij that represents the error term. Thus, we estimate the coefficients β, γ,
and λ where the latter two are vectors.
The data come from the World Value Survey (WVS), which is a worldwide survey
carried out by the Inter-University Consortium for Political and Social Research (ICPSR) that
comprises individual cross-national questions on a wide variety of topics, such as the economy,
politics, foreign policy, and identity, as well as on the socio-economic background of individual
respondents and his or her attitudes on several topics. Data come from in-person interviews of a
sampling universe of adult citizens 15 years old and older from different developed and
developing countries around the world. For our purposes, as we need specific variables related to
democracy, our sample is composed of around 240,000 individuals (from 85 countries) that were
surveyed during the so-called third, fourth and fifth waves, between 1994 and 2008.8
8
The countries along with their respective samples and the wave in which was executed the survey are presented in
Appendix 1.
11
As measures of our dependent variable, democracy, we use a broad set of proxies. The
first one is included in the three waves and is regarded as our preferred measure. It answers the
question: “Would you say having a democratic political system is a very good, fairly good, fairly
bad or very bad way of governing this country?” The other three proxies considered are included
in the third and fourth waves only, showing individuals’ agreement, on a scale of 1 to 4, with the
following statements: (i) “In democracy economic system does not run badly”; (ii) “Democracies
are good at maintaining order”; and (iii) “Democracy may have problems but is better”. For the
sake of completeness we use another proxy that is included in the fifth wave only. On a scale
from 1 to 10, this variable answers the question: “How important is it for you to live in a country
that is governed democratically?” The five proxies used are described in detail in Table 1. They
are all categorical variables, and, consequently, the coefficients are estimated using ordered
probits.9
As mentioned above, the most important explanatory variable is years of schooling of
individuals. We transform the available categorical variable into pseudo years of education
according to each level attained. Along with our variable of interest, the other explanatory
variables included in the benchmark specification are the following: age, a dummy variable for
gender that equals one for women, a dummy variable for marital status that equals one for being
single, two dummy variables for employment status: employed, and unemployed; a dummy
variable that equals one when the individuals live with their parents, and a scale of income on
which the individuals’ household is located according to their perception.10
We also use other control variables that are not included in the benchmark specification
because they significantly reduce the sample. These variables are the following: a categorical
variable that represents the size of the town of residence, the perceived degree of respect for
individual human rights in the respondent’s country, and self-positioning on a political scale
(from left to right). In other regressions, we add to the benchmark specification variables related
to individuals’ satisfaction with democracy in their own countries. One variable directly
addresses dissatisfaction with the development of democracy in the respondent’s country (third
and fourth wave), and the other asks how democratically the respondent’s country is being
9
Ordinary least squares were also estimated, and results are analogous.
For more details about the definition of these and the following variables presented in this section, see Table 1.
10
12
governed (fifth wave). Summary statistics of these variables and those described in the preceding
paragraph are presented in Table 2.
Additionally, all estimations include country and year dummies, have robust standard
errors, and are clustered by the country and wave in which the survey was executed. Table 3
presents a correlation matrix between all our proxies of Democracy and the explanatory
variables, with corresponding p-values of the coefficients.
Additionally, for illustrative purposes Figure 1 presents the average years of schooling of
people that approve and disapprove democracy, using our five proxies, by region (as classified
by the World Bank). It appears that, across regions, less educated people favor democracy less
than do more educated people.
4. Regression Results
In Table 4, we present the results of ordered probit regressions using the approval of a
democratic system as our dependent variable. The first column presents our main results using
the benchmark specification, and the other columns use the additional explanatory variables as
explained above. In the four regressions presented, schooling yields a positive statistically
significant coefficient at one percent. Additionally, being older and male, whether employed or
unemployed (compared to students, retired and housewives) seems to have a robust negative and
statistically significant effect on the dependent variable. In the second column, results show that
residents of large towns and individuals maintaining that human rights are respected in their
country are more likely to approve of democracy.
The corresponding marginal effects of these coefficients are presented in Table 5. For the
sake of economy we only show the change of an average individual in the sample who believes
that it is fairly good to have a democratic system.11 In the first column, one more year of
schooling implies that the probability of perceiving a democratic system as good (not just fairly
good) is estimated to increase by 1.2 percentage points. This magnitude is very similar in the
other columns, with an increase of between 1.1 and 1.4 percentage points.
In the last four columns of Table 4 we evaluate the effects of the perception of democracy
and its interaction with education on the preference for democracy. Individuals not satisfied with
11
Summary statistics presented in Table 2 show that the mean of the variable “Approval of a democratic political
system” is 3.
13
the way democracy develops in their country approve less of democracy (columns 3 and 4), and
those who think that their government is more democratic approve of it more than otherwise
(columns 5 and 6). We also add interactive terms between education and democracy, and they
yield coefficients that are not statistically significant at conventional levels. This result implies
that the effect on pro-democracy views is positive independently of the type of regime. This is
fully consistent with our theoretical model that argues that the informative benefits of education
are universal across regimes and is not consistent with the implication of Lott (1990) implication
that totalitarian regimes successfully use education to indoctrinate citizens into supporting the
regime.
To further address this issue, in Table 6 we provide results based on dividing the sample
into two sub-samples, employing the countries’ average democracy level to this end. In
particular, in order to estimate the cut-off point used to divide the sample, we use the mean of the
Democratic Accountability variable of the International Country Risk Guide (ICRG). Unlike our
previously used proxy, this variable is available at the country level only. However, it is not
based on the perception on the individual, but on an exogenous (albeit still subjective)
assessment of the extent to which a country is democratic.12 Our results show that in both types
of countries, the more democratic and the less democratic, the effects of education on the
preference for democracy are statistically significant, although its coefficient is somewhat
smaller in the latter case. In Appendix 2, we execute the same exercise for a smaller subsample
of the least democratic countries, and the results still hold, although the education coefficient
further drops, indicating that education predicts pro-democratic attitudes in non-democracies, but
somewhat less well than in democracies.13
The coefficients resulting from the regressions using the other four Democracy proxies
available are presented in Table 7, and the marginal effects are shown in Appendix 4. The
coefficients for our variable of interest are again positive and statistically significant at one
percent. With respect to the other explanatory variables, we find that older individuals tend to
agree with the idea that democracy may have problems but is better than the alternative (column
3) and that living in a country governed democratically is important (column 4); but they tend to
12
Using the median instead of the mean does not change our findings. Similarly, we also used Polity V instead of
ICRG and our results are qualitatively identical. These additional results are available upon request.
13
Appendix 3 shows the results of dividing the sample at the individual level, by using the same democracy variable
employed in the interactions presented in Tables 4 and 5. The results are very similar.
14
disagree with the idea that democracies are good at maintaining order (column 2). As in Table 4,
being a male or belonging to a higher income level has a positive and statistically significant
effect on pro-democracy preferences.
We find that the coefficient of the town size is positive and statistically significant in two
of the four regressions presented in Table 7. Living in countries where individuals perceive that
there is more respect of human rights, or that democracy is sufficiently developed, again has a
statistically significant and positive effect.
Endogeneity may be an issue of concern, primarily because of potentially omitted
variables. (Notice that, as we deal with democratic attitudes, reverse causality is unlikely here.)
To address this issue, Tables 8 and 9 present our findings using instrumental variables (IV). In
both tables we use (i) the savings of the respondent’s family during the previous year and (ii) the
respondent’s number of children as instrumental variables. These instruments appear to be
sensitive. Income and thus savings are typically positively correlated with education; however,
there is no reason to expect that level of savings may be linked in a consistent way with
democratic perceptions. Similarly, while one would expect a negative correlation between the
number of children and the individual’s education level, the former is unlikely to be directly
correlated with democratic perceptions.14
In Table 8 we present the benchmark specification of Table 4 as well as an additional
specification that includes additional explanatory variables (but reduces the sample size
dramatically). This table also includes the marginal effects of a change from the mean, in this
case, the probability increase of being in category 4 (thinking that having a democratic political
system is a very good way of governing the country). In both regressions the effect of schooling
on preference for democracy is the same as before: a coefficient that is positive and statistically
significant at conventional levels. The marginal effect of education is somewhat smaller than the
one without instrumental variables. The probability of the belief that having a democratic
political system is good, not just fairly good, increases by 1.3 and 1.2 percentage points, in
columns 2 and 4, respectively. From the last row of this table, which presents the corresponding
p-value significance of both instruments, it is seen that they are statistically significant.15
14
15
China may be considered an outlier. When excluding this country from our sample our results do not change.
Also, corresponding tests of exclusion restrictions show that the instruments employed are good ones.
15
In Table 9 we present the results of the benchmark specification using the other four
proxies of Democracy. We also include in this table the marginal effects. The education
coefficient is positive and statistically significant, and the marginal effects are smaller but also
result statistically significant. The first stage regressions of Table 8 and 9 are presented in
Appendix 5. In those regressions the corresponding coefficients have the expected signs: less
educated individuals have a larger number of children than more educated ones; and families of
less educated individuals just get by or spend savings.
5. Concluding Remarks
The relationship between education and democracy has been a long standing subject of interest,
both theoretically and empirically.
Influential commentators have suggested mechanisms
through which education may promote democracy, and the link between the two has been well
documented in the recent literature. Discerning the causal impact of education, however, has
proved difficult because of endogeneity issues.
Further, suggestions have been made that
education may promote democracy under democratic regimes only, whereas in non-democratic
setting the effect might reverse itself.
In this paper, therefore, we attempt to make a twofold contribution. Using rich survey
data, we study whether education promotes pro-democratic preferences; and whether this effect
differs across countries with different levels of democracy. Our theoretical model, formalizing
the sentiment that education confers informational advantages, essential for democracy, suggests
that educated individuals have a comparative advantage in a democracy in translating their
(better) judgment of public policies into action. This informational advantage emerges even
when the ruler uses deception to misinform the population in order to tilt outcomes to his own
advantage. Consequently, the educated favor democracy more than less educated individuals.
The empirical part tests the main implications of our analysis using individual level data
from across the globe. We find a positive effect of education on pro-democratic attitudes, using
a variety of proxies for the latter, personal characteristics as controls, and also addressing
endogeneity issues.
For example, one more year of schooling implies an increase in the
likelihood of perceiving a democratic system as good, as opposed to fairly good, by more than
one percentage point. Moreover, the effect exists regardless of a country’s level of democracy,
16
suggesting that the possible use of education for indoctrination purposes in non-democratic
environments may have limited consequences.
Based on survey data, this research complements previous efforts that find a positive
effect of education on measures of civic participation. Whereas the latter are often interpreted as
being correlated with democracy, the relationship is not straightforward, as one could imagine
social activities whose nature is precisely anti-democratic. In this sense, findings presented here
offer yet another, and one that is perhaps more direct, on how education may affect democracy.
Additionally, this study questions the view that education promotes democratic views only under
democracies, as the effect identified here appears to hold universally. We view this second result
as casting doubt on the indoctrinating potential of education in non-democratic settings.
In future work, it would be interesting to disintegrate the analysis by the respondent’s
occupation and sector of employment. This has the potential of relating the link between
education and democratic preferences to the respondents’ affiliation with the governing regime.
For instance, bureaucrats, army officers, and governing party members in non-democratic
environments may exhibit preference that depend more on their occupation and employment than
on education.
17
References
Acemoglu, D. et al. 2005. “From Education to Democracy.” American Economic Review 95: 4449.
Barro, R.J. 1999. “The Determinants of Democracy.” Journal of Political Economy 107: S158S183.
Bobba, M., and D. Coviello. 2007. “Weak Instruments and Weak Identification in Estimating
the Effects of Education on Democracy.” Economic Letters 96: 301-307.
Bourguignon, F., and T. Verdier. 2000. “Oligarchy, Democracy, Inequality and Growth.”
Journal of Development Economics 62: 285-313.
Brady, H.E., S. Verba and K. L. Schlozman. 1995. “Beyond SES: A Resource Model of Political
Participation.” American Political Science Review 89: 271-294.
Campante, F., and D. Chor. 2008. “Schooling and Political Participation in a Neoclassical
Framework,” mimeo.
Dee, T. 2004. “Are There Civic Returns to Education?” Journal of Public Economics 88: 16971720.
Dewey, J. 1916. Democracy and Education. New York, United States: The Macmillan
Company.
Friedman, M. 1962. Capitalism and Freedom. Chicago, United States: University of Chicago
Press.
Glaeser, E. et al. 2004. “Do Institutions Cause Growth?” Journal of Economic Growth 9: 271303.
Glaeser, E., G. Ponzetto, and A. Shleifer. 2007. “Why Does Democracy Need Education?”
Journal of Economic Growth 12: 77-99.
Kartik, N., M. Ottaviani and F. Squintani. 2007. “Credulity, Lies, and Costly Talk.” Journal of
Economic Theory 134: 93-116.
La Porta, R. et al. 1999. “The Quality of Government.” Journal of Law, Economics and
Organization 15: 222-279.
Lipset, S. 1959. “Some Social Requisites of Democracy: Economic Development and Political
Legitimacy.” American Political Science Review 53: 69-105.
Lott, J.R. Jr., 1990. “An Explanation for Public Provision of Schooling: The Importance of
Indoctrination.” Journal of Law and Economics 33: 199-232.
18
----. 1999. “Public Schooling, Indoctrination, and Totalitarianism.” Journal of Political Economy
107(6/Part 2): S127-S157.
Mann, H. 1846. Tenth Annual Report of the Secretary of the Massachusetts State Board of
Education. Boston: Horace B. Fuller.
Milligan, K., E. Moretti, and P. Oreopoulos. 2004. “Does Education Improve Citizenship?
Evidence from the United States and the United Kingdom.” Journal of Public Economics
88: 1667-1695.
Spilimbergo, A. 2008. “Democracy and Foreign Students.” London, United Kingdom: Centre for
Economic Policy Research. Available at http://antoniospilimbergo.com/research.htm.
Forthcoming in American Economic Review.
19
Figure 1. Schooling and Approval of Democracy
Agreement with "In democracy, economic system does not run badly"
12
11
10
9
8
7
6
5
Years of schooling
Years of schooling
Approval of a democratic political system
High income: OECD
High Low income Lower income: middle nonOECD
income
"Having a democratic political system" is
Very bad 1
13
12
11
10
9
8
7
6
5
High income: OECD
Upper middle income
Very good 4
Strongly disagree 1
Years of schooling
High Low income Lower income: middle nonOECD
income
Strongly disagree 1
Upper middle income
Strongly agree 4
13
12
11
10
9
8
7
6
5
High income: OECD
Strongly agree 4
High Low income Lower income: middle nonOECD
income
Strongly disagree 1
Importance of living in a country that is governed democratically
14
12
10
8
6
4
2
0
High income: OECD
Upper middle income
Agreement with "D emocracy may have problems but is better"
13
12
11
10
9
8
7
6
5
Years of schooling
Years of schooling
Agreement with "Democracies are good at maintaining order"
High income: OECD
High Low income Lower income: middle nonOECD
income
High Low income Lower income: middle nonOECD
income
Not at all important 1
20
Upper middle income
Absolutely important 10
Strongly agree 4
Upper middle income
Table 1. Variables Description
Variable
Description
Individual Level Variables
Approval of a democratic
The question in the survey is as follows: Would you say it is a (1) very good, (2) fairly
political system
good, (3) bad or (4) very bad way of governing this country having a democratic political
system. This variable scale was changed to: (1) very bad – (4) very good. Source: World
Value Surveys (Third, Fourth and Fifth Waves).
Agreement with opinion that The question in the survey is: Could you please tell me if you (1) agree strongly, (2) agree,
says that in democracy
(3) disagree or (4) strongly disagree with this statement “In democracy economic system
economic system does not runs badly”. Source: World Value Surveys (Third and Fourth Waves).
run badly
Agreement with opinion that The question in the survey is: Could you please tell me if you (1) agree strongly, (2) agree,
says that democracies are
(3) disagree or (4) strongly disagree with this statement “Democracies aren’t good at
good at maintaining order maintaining order”. Source: World Value Surveys (Third and Fourth Waves).
Agreement with the opinion
that says that democracy
may have problems but is
better
Importance of living in a
country that is governed
democratically
Schooling
Age
Gender
Marital status
Employment status
Scale of income
Size of town
The question in the survey is: Could you please tell me if you (1) agree strongly, (2) agree,
(3) disagree or (4) strongly disagree whit this statement “Democracy may have problems
but is better than any other form of government”. This variable scale was changed to: (1)
strongly disagree – (4) agree strongly. Source: World Value Surveys (Third and Fourth
Waves).
The question in the survey is: How important is it for you to live in a country that is
governed democratically? On this scale where 1 means “it is not important at all” and 10
means “absolutely important”. Source: World Value Surveys (Fifth Wave).
The exact question in the survey is: What is the highest educational level that you have
attained? (1) Inadequately completed elementary education, (2) Completed (compulsory)
elementary education, (3) Incomplete secondary school: technical/vocational type, (4)
Complete secondary school: technical/vocational type, (5) Incomplete secondary:
university-preparatory, (6) Complete secondary: university-preparatory, (7) Some
university without degree/higher education, (8) University with degree/higher education.
This variable was changed to one with pseudo years of education, according to each level:
To (1) we assigned 3 years of schooling; to (2), 6; to (3), 8.5; to (4), 11; to (5), 12.5; to (6),
14; to (7), 13.5; and to (8), 16. Source: World Value Surveys (Third, Fourth and Fifth
Waves).
Respondent’s age. Source: World Value Surveys (Third, Fourth and Fifth Waves).
Gender of the respondent. (1) Female and (0) Male. Source: World Value Surveys (Third,
Fourth and Fifth Waves).
2 dummies: (1) Married, living together as married, divorced, separated or widowed, and
(2) Single. In all the regressions (1) is the omitted dummy. Source: World Value Surveys
(Third, Fourth and Fifth Waves).
Employment status composed of 3 dummies: (1) Employed (Part or full time) and Selfemployed; (2) Retired/pensioned, Housewife not otherwise employed, and Student; and (3)
Unemployed. In all the regressions, dummy (2) is omitted. Source: World Value Surveys
(Third, Fourth and Fifth Waves).
A scale of incomes in which the household falls into, before taxes and other deductions.
This variable takes values from 1 to 10, 1 being the lowest decile and 10 the highest. The
data is recollected in local currency, scaled and then aggregated so the deciles represent a
country level income ranking. Source: World Value Surveys (Third, Fourth and Fifth
Waves).
Categorical variable: (1) Under 2,000; (2) 2-5,000; (3) 5-10,000; (4) 10-20,000; (5) 2050,000; (6) 50-100,000; (7) 100-500,000; and (8) 500,000 and more. Source: World Value
Surveys (Third, Fourth and Fifth Waves).
21
Variable
Description
Respect of human rights in The question in the survey is: How much respect is there for individual human rights
own country
nowadays in this country? Do you feel there is: (1) A great deal of respect for individual
rights, (2) Fairly much respect, (3) Not much respect, and (4) No respect at all. This
variable scale was changed to: (1) No respect at all – (4) A great deal of respect (...).
Source: World Value Surveys (Third, Fourth and Fifth Waves).
Self positioning in political How the respondent place his/her views on the scale from (1) Left to (10) Right. Source:
scale
World Value Surveys (Third, Fourth and Fifth Waves).
No satisfaction with the way The question in the survey is: On the whole are you very satisfied, not very satisfied or not
democracy develops
at all satisfied with the way democracy is developing in our country? (1) Very satisfied, (2)
Rather satisfied, (3) Not very satisfied, and (4) Not at all satisfied. Source: World Value
Surveys (Third and Fourth Waves).
Democracy in own country The question in the survey is: How democratically is this country being governed today?
Again using a scale from 1 to 10, where 1 means it is “not at all democratic” and 10 means
that it is “completely democratic”, what position you use? Source: World Value Surveys
(Fifth Wave).
Number of children
Number of children, where 0 means no children. Source: World Value Surveys (Third,
Fourth and Fifth Waves).
Savings
4 dummies that answer to the question: During the past year, did your family: (1) Save
money, (2) Just get by, (3) Spent some savings, and (4) Spent savings and borrowed
money. The omitted category is the first one. Source: World Value Surveys (Third, Fourth
and Fifth Waves).
Country Level Variables
Democratic Accountability This is a measure of how responsive government is to its people, on the basis that the less
responsive it is, the more likely it is that the government will fall, peacefully in a
democratic society, but possibly violently in a non-democratic one. Average for five years,
including the year when the individual was surveyed. Source: ICRG.
22
Table 2. Summary Statistics
Variable
Obs.
Mean
Min
Max
3.3660
2.7430
Std.
Dev.
0.7219
0.7914
Approval of a democratic political system
Agreement with opinion that says that in
democracy economic system does not run badly
Agreement with opinion that says that democracies
are good at maintaining order
168,542
99,476
1
1
4
4
101,648
2.7352
0.8124
1
4
103,346
3.2669
0.7353
1
4
Agreement with the opinion that says that
democracy may have problems but is better
Importance of living in a country that is governed
democratically
Schooling
Age
Gender: Female
Marital Status: Single
Living with her parents
Employment status: Employed
Employment status: Unemployed
Scale of income
Size of town
Respect of human rights in own country
49,204
8.6028
1.9138
1
10
168,542
168,542
168,542
168,542
168,542
168,542
168,542
168,542
118,529
125,978
10.6126
40.7457
0.5026
0.2485
0.2727
0.5428
0.0924
4.6327
4.9526
2.5779
4.1992
15.8711
0.5000
0.4321
0.4453
0.4982
0.2896
2.4031
2.4838
0.8752
3
15
0
0
0
0
0
1
1
1
16
99
1
1
1
1
1
10
8
4
Self-positioning in political scale
Satisfaction with the way democracy develops
Democracy in own country
Savings: Family saved money
Savings: Family just got by
Savings: Family spent some savings
129,160
71,819
45,647
132,582
132,582
132,582
5.6894
2.4111
6.4195
0.2388
0.4910
0.1469
2.3777
0.8322
2.4270
0.4264
0.4999
0.3540
1
1
1
0
0
0
10
4
10
1
1
1
Savings: Family spent savings and borrowed
money
Number of children
132,582
0.1234
0.3288
0
1
162,544
1.9154
1.7958
0
20
23
Table 3. Correlation between Individual-Level Variables and Preference for Democracy
Schooling
Age
Gender: Female
Marital Status: Single
Living with her parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
Size of town
Respect of human rights in
own country
Self-positioning in political
scale
Satisfaction with the way
democracy develops
Democracy in own country
Approval of
a
democratic
political
system
Agreement
with
opinion that
says that in
democracy
economic
system does
not run
badly
Agreement with
opinion that says
that democracies
are good at
maintaining
order
Agreement
with the
opinion that
says that
democracy
may have
problems but
is better
Importance of
living in a
country that is
governed
democratically
0.038
0.000
-0.0067
0.006
-0.0386
0.000
0.0239
0.000
0.0074
0.002
0.0164
0.000
-0.0174
0.000
0.0693
0.000
0.0173
0.000
0.1371
0.000
0.0041
0.138
0.2169
0.000
0.1024
0.000
0.0857
0.000
-0.0097
0.002
-0.0437
0.000
0.0168
0.000
-0.017
0.000
0.0491
0.000
-0.0423
0.000
0.1213
0.000
0.0185
0.000
0.1282
0.000
0.0115
0.001
0.1581
0.000
-
0.0802
0.000
-0.0393
0.000
-0.026
0.000
0.028
0.000
-0.0027
0.382
0.047
0.000
-0.0245
0.000
0.1033
0.000
0.0206
0.000
0.1389
0.000
-0.006
0.087
0.1657
0.000
-
0.026
0.000
0.0342
0.000
-0.0254
0.000
-0.0085
0.006
-0.0159
0.000
0.014
0.000
-0.0228
0.000
0.0508
0.000
0.0138
0.000
0.1208
0.000
0.0296
0.000
0.1632
0.000
-
0.0733
0.000
0.0521
0.000
-0.0111
0.013
-0.0111
0.014
-0.0534
0.000
0.0145
0.001
-0.035
0.000
0.0624
0.000
0.0654
0.000
0.0988
0.000
0.0389
0.000
-
24
0.2332
0.000
Table 4. Schooling and Approval of a Democratic Political System (Coefficients)
Schooling
Age
Gender: Female
Marital Status: Single
Living with her parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
Size of town
Respect of human rights in
own country
Self-positioning in political
scale
Satisfaction with the way
democracy develops
Satisfaction with the way
democracy develops *
Schooling
Democracy in own country
Ordered Probit Regressions (Coefficients)
Dependent variable: Approval of a democratic political system
(1)
(2)
(3)
(4)
(5)
(6)
0.030*** 0.036*** 0.035*** 0.048*** 0.026***
0.020***
(0.002)
(0.003)
(0.004)
(0.014)
(0.004)
(0.006)
0.004*** 0.004*** 0.003*** 0.003*** 0.004***
0.004***
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
-0.068*** -0.069*** -0.081*** -0.082*** -0.044*** -0.043***
(0.009)
(0.012)
(0.015)
(0.015)
(0.014)
(0.014)
0.023*
0.022
0.017
0.017
0.008
0.008
(0.012)
(0.017)
(0.017)
(0.017)
(0.022)
(0.022)
-0.007
-0.023
-0.009
-0.009
0.005
0.005
(0.013)
(0.018)
(0.020)
(0.021)
(0.024)
(0.024)
-0.008
-0.043*** -0.031**
-0.032**
-0.016
-0.016
(0.011)
(0.016)
(0.016)
(0.016)
(0.020)
(0.020)
-0.042** -0.076*** -0.084*** -0.084***
-0.000
-0.000
(0.017)
(0.023)
(0.023)
(0.023)
(0.027)
(0.027)
0.017*** 0.017*** 0.017*** 0.017***
0.005
0.005
(0.004)
(0.004)
(0.004)
(0.004)
(0.007)
(0.007)
0.009**
(0.004)
0.153***
(0.019)
0.006
(0.007)
0.213*** 0.271***
(0.027)
(0.067)
-0.006
(0.005)
0.049***
(0.009)
Democracy in own country *
Schooling
Country dummies
Region dummies
Year dummies
Waves
Observations
Number of Country Waves
Yes
No
Yes
3, 4, 5
168,542
140
Yes
No
Yes
3, 4, 5
66,661
80
Yes
No
Yes
3,4
71,819
67
Yes
No
Yes
3,4
71,819
67
Yes
No
Yes
5
45,647
37
0.038***
(0.014)
0.001
(0.001)
Yes
No
Yes
5
45,647
37
(clusters)
Pseudo R2
0.0642
0.0776
0.0780
0.0781
0.0581
0.0581
Robust standard errors are presented in parentheses. They are clustered by country and corresponding survey wave.
The row called “Waves” specifies the survey waves included in the sample. In column (1), for example, the
regression only includes the 3rd, 4th and 5th waves of the survey. See appendix 1 for correspondence of survey wave
and year of survey. *** p<0.01, ** p<0.05, * p<0.1
25
Table 5. Schooling and Approval of a Democratic Political System (Marginal Effects)
Ordered Probit Regressions (Marginal Effects, =4)
Dependent variable: Approval of a democratic political system
(1)
(2)
(3)
(5)
Schooling
0.012***
0.014***
0.014***
0.011***
(0.001)
(0.001)
(0.002)
(0.001)
Age
0.002***
0.002***
0.001***
0.002***
(0.000)
(0.000)
(0.000)
(0.000)
Gender: Female
-0.027***
-0.027***
-0.032***
-0.017***
(0.003)
(0.005)
(0.006)
(0.006)
Marital Status: Single
0.009*
0.009
0.007
0.003
(0.005)
(0.007)
(0.007)
(0.009)
Living with her parents
-0.003
-0.009
-0.004
0.002
(0.005)
(0.007)
(0.008)
(0.010)
Employment status: Employed
-0.003
-0.017***
-0.013**
-0.006
(0.004)
(0.006)
(0.006)
(0.008)
Employment status: Unemployed
-0.017**
-0.030***
-0.033***
-0.000
(0.007)
(0.009)
(0.009)
(0.011)
Scale of income
0.007***
0.007***
0.007***
0.002
(0.001)
(0.002)
(0.002)
(0.003)
Size of town
0.004**
(0.002)
Respect of human rights in own country
0.061***
(0.008)
Self-positioning in political scale
0.002
(0.003)
Satisfaction with the way democracy
0.085***
develops
(0.011)
Democracy in own country
0.019***
(0.004)
3, 4, 5
3, 4, 5
3, 4
5
Waves (clusters)
Observations
168,542
66,661
71,819
45,647
Marginal effects measure the change in the probability of choosing the fourth category. Column (4) corresponds to
the marginal effects of column (5) in Table 4. Robust standard errors are presented in parentheses. They are
clustered by country and corresponding survey wave. The row called “Waves” specifies the survey waves included
in the sample. In column (1), for example, the regression only includes the 3rd, 4th and 5th waves of the survey. See
appendix 1 for correspondence of survey wave and year of survey. *** p<0.01, ** p<0.05, * p<0.1
26
Table 6. Education and Measures of Preference for Democracy by sub-samples
Ordered Probit Regressions
Dependent variable: Approval of a democratic political system
ICRG - Democratic Accountability
ICRG - Democratic
<= mean (4.3)
Accountability > mean (4.3)
(1)
Coefficient
(2)
Marginal Effect (=4)
Schooling
(3)
Coefficient
(4)
Marginal Effect
(=4)
0.021***
0.008***
0.036***
0.014***
(0.003)
(0.001)
(0.004)
(0.002)
Age
0.003***
0.001***
0.006***
0.002***
(0.001)
(0.000)
(0.001)
(0.000)
Gender: Female
-0.041***
-0.016***
-0.084***
-0.033***
(0.014)
(0.006)
(0.012)
(0.005)
Marital Status: Single
0.017
0.007
0.030
0.012
(0.015)
(0.006)
(0.018)
(0.007)
Living with her parents
-0.008
-0.003
-0.003
-0.001
(0.020)
(0.008)
(0.018)
(0.007)
Employment status: Employed
0.014
0.006
-0.011
-0.005
(0.017)
(0.007)
(0.015)
(0.006)
Employment status:
-0.015
-0.006
-0.048*
-0.019*
Unemployed
(0.018)
(0.007)
(0.029)
(0.012)
Scale of income
0.017***
0.007***
0.021***
0.008***
(0.006)
(0.002)
(0.005)
(0.002)
Country dummies
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Observations
68,149
68,149
88,025
88,025
Country Waves (clusters)
51
80
Pseudo R2
0.0758
0.0558
To estimate the sub-samples, we use the mean value of the variable “Democratic Accountability Index” (from
ICRG). The cut-off point is shown in the table; corresponding definitions of variables are shown in Table 1.
Marginal effects measure the change in the probability of choosing the fourth category. Robust standard errors are
presented in parentheses. They are clustered by country and wave. The sample in all these regressions includes three
waves (3rd, 4th and 5th). See appendix 1 for correspondence of survey wave and year of survey. *** p<0.01, **
p<0.05, * p<0.1.
27
Table 7. Education and Measures of Preference for Democracy (Coefficients)
Agree with "In democracy,
economic system does not
run badly"
Schooling
Age
Gender: Female
Marital Status: Single
Living with her
parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
(1)
0.027***
(0.003)
0.001
(0.001)
-0.086***
(0.011)
0.040**
(0.016)
-0.044***
(0.015)
0.010
(0.011)
-0.076***
(0.017)
0.032***
(0.004)
Size of town
Respect of human
rights in own country
Self-positioning in
political scale
Satisfaction with the
way democracy
develops
Democracy in own
country
Country dummies
Year dummies
Waves
Observations
Number of Country
Waves (clusters)
Pseudo R2
(2)
0.033***
(0.004)
0.000
(0.001)
-0.092***
(0.014)
0.035*
(0.020)
-0.057***
(0.021)
0.003
(0.018)
-0.079**
(0.031)
0.032***
(0.004)
0.005
(0.005)
0.090***
(0.022)
0.014*
(0.007)
0.135***
(0.026)
Ordered Probit Regressions (Coefficients)
Dependent variable:
Agree with "Democracies
Agree with "Democracy
are good at maintaining
may have problems but is
order"
better"
(3)
0.027***
(0.003)
-0.001***
(0.000)
-0.044***
(0.009)
0.018
(0.013)
-0.039***
(0.012)
0.016
(0.010)
-0.038**
(0.016)
0.026***
(0.004)
(4)
0.031***
(0.004)
-0.002***
(0.001)
-0.031**
(0.013)
-0.001
(0.019)
-0.045**
(0.018)
0.020
(0.015)
-0.048*
(0.028)
0.023***
(0.004)
0.010**
(0.005)
0.083***
(0.020)
-0.004
(0.009)
0.123***
(0.021)
(5)
0.027***
(0.003)
0.004***
(0.001)
-0.046***
(0.010)
-0.014
(0.014)
0.008
(0.014)
0.002
(0.012)
-0.029
(0.022)
0.018***
(0.005)
(6)
0.030***
(0.006)
0.004***
(0.001)
-0.049***
(0.013)
-0.035
(0.022)
0.025
(0.020)
-0.017
(0.017)
-0.047
(0.032)
0.018***
(0.005)
0.012***
(0.004)
0.100***
(0.022)
0.020**
(0.010)
0.156***
(0.027)
Importance of living in
a country that is
governed
democratically
(7)
(8)
0.031***
0.039***
(0.003)
(0.005)
0.005***
0.006***
(0.001)
(0.002)
-0.030*
-0.033
(0.016)
(0.025)
-0.003
-0.019
(0.017)
(0.031)
0.000
-0.003
(0.016)
(0.024)
0.003
-0.014
(0.018)
(0.028)
-0.023
-0.071
(0.036)
(0.054)
0.008
-0.003
(0.008)
(0.010)
0.014
(0.009)
0.028
(0.031)
0.018
(0.013)
0.099***
(0.015)
Yes
Yes
3,4
105,417
97
Yes
Yes
3,4
42,415
54
Yes
Yes
3,4
107,692
98
Yes
Yes
3,4
42,914
54
Yes
Yes
3,4
108,941
97
Yes
Yes
3,4
43,520
54
Yes
Yes
5
52,766
39
Yes
Yes
5
23,310
26
0.0433
0.0548
0.0424
0.0595
0.0621
0.0966
0.0295
0.0436
Robust standard errors are presented in parentheses. They are clustered by country and corresponding survey wave. The row
called “Waves” specifies the survey waves included in the sample. In column (1), for example, the regression only includes the
3rd and 4th waves of the survey. See appendix 1 for correspondence of survey wave and year of survey. *** p<0.01, ** p<0.05, *
p<0.1.
28
Table 8. Education and Approval of a Democratic Political System, IV estimates
Ordered Probit Regressions and Instrumental Variables
Dependent variable: Approval of a democratic political system
(1)
(2)
(3)
(4)
Coefficient
Marginal
Coefficient
Marginal Effect
Effect (=4)
(=10)
Schooling
Age
Gender: Female
Marital Status: Single
Living with her parents
Employment status: Employed
Employment status: Unemployed
Scale of income
0.032***
(0.007)
0.004***
(0.001)
-0.064***
(0.011)
0.023
(0.015)
-0.006
(0.015)
-0.013
(0.015)
-0.042**
(0.019)
0.011**
(0.005)
Size of town
Respect of human rights in own country
Self-positioning in political scale
0.013***
(0.003)
0.002***
(0.000)
-0.026***
(0.004)
0.009
(0.006)
-0.002
(0.006)
-0.005
(0.006)
-0.017**
(0.008)
0.004**
(0.002)
0.030*
(0.016)
0.004**
(0.002)
-0.064***
(0.017)
0.032
(0.025)
-0.029
(0.023)
-0.056*
(0.029)
-0.110***
(0.030)
0.010
(0.009)
0.007
(0.006)
0.102***
(0.023)
-0.000
(0.008)
No
Yes
3, 4, 5
39,224
50
0.0634
0.012*
(0.006)
0.002**
(0.001)
-0.025***
(0.007)
0.013
(0.010)
-0.011
(0.009)
-0.022*
(0.012)
-0.044***
(0.012)
0.004
(0.004)
0.003
(0.002)
0.041***
(0.009)
-0.000
(0.003)
No
Yes
3, 4, 5
39,224
Country dummies
Yes
Yes
Year dummies
Yes
Yes
Waves
3, 4, 5
3, 4, 5
Observations
129,506
129,506
112
Number of Country Waves (clusters)
Pseudo R2
0.0576
Instruments Significance
0.000
0.000
(1st stage, P-value)
Robust standard errors are presented in parentheses. They are clustered by country and corresponding survey wave.
These are the results of the second stage regressions. In the first stages, education is explained by all the other
controls presented here and the excluded instruments: family savings and number of children. The results of the first
stages are presented in Appendix 4. Marginal effects measure the change in the probability of choosing the fourth
category. The row called “Waves” specifies the survey waves included in the sample. In column (1), for example,
the regression only includes the 3rd and 4th waves of the survey. See appendix 1 for correspondence of survey wave
and year of survey. *** p<0.01, ** p<0.05, * p<0.1
29
Table 9. Education and Measures of Preference for Democracy IV Estimates
Agree with "In
democracy, economic
system does not run
badly"
(1)
Coefficient
Schooling
Age
Gender: Female
Marital Status:
Single
Living with her
parents
Employment
status: Employed
Employment
status:
Unemployed
Scale of income
Country dummies
Year dummies
Waves
Observations
Number of
Country Waves
(clusters)
Pseudo R2
Instruments
Signficance (1st
stage, P-value)
0.019*
(0.011)
0.000
(0.001)
-0.080***
(0.013)
0.049**
(0.020)
-0.038**
(0.017)
0.016
(0.015)
-0.067***
(0.019)
0.030***
(0.007)
Yes
Yes
3, 4
74,299
71
Ordered Probit Regressions and Instrumental Variables
Dependent variable:
Agree with
Agree with "Democracy
Importance of living in
"Democracies are good
may have problems but
a country that is
at maintaining order"
is better"
governed
democratically
(2)
Marginal
Effect
(=4)
0.004*
(0.003)
0.000
(0.000)
-0.018***
(0.003)
0.011**
(0.004)
-0.008**
(0.004)
0.004
(0.003)
-0.015***
(3)
Coefficient
(0.004)
0.007***
(0.001)
Yes
Yes
3, 4
74,299
(0.019)
0.022***
(0.006)
No
Yes
3, 4
75,837
72
0.023***
(0.009)
-0.001
(0.001)
-0.055***
(0.011)
0.029*
(0.015)
-0.031**
(0.015)
0.012
(0.013)
-0.040**
(4)
Marginal
Effect
(=4)
0.005***
(0.002)
-0.000
(0.000)
-0.013***
(0.002)
0.007*
(0.004)
-0.007**
(0.003)
0.003
(0.003)
-0.009**
(5)
Coefficient
(0.004)
0.005***
(0.001)
No
Yes
3, 4
75,837
(0.023)
0.016**
(0.006)
Yes
Yes
3, 4
76,248
71
0.020*
(0.010)
0.004***
(0.001)
-0.036***
(0.013)
-0.010
(0.019)
0.017
(0.017)
-0.006
(0.016)
-0.017
(6)
Marginal
Effect
(=4)
0.008*
(0.004)
0.002***
(0.000)
-0.014***
(0.005)
-0.004
(0.007)
0.007
(0.006)
-0.002
(0.006)
-0.006
(7)
Coefficient
(0.009)
0.006**
(0.003)
Yes
Yes
3, 4
76,248
(0.039)
0.004
(0.010)
No
Yes
5
46,795
37
0.036***
(0.013)
0.005***
(0.002)
-0.032*
(0.018)
-0.018
(0.022)
-0.011
(0.018)
0.006
(0.024)
-0.017
0.0395
0.0334
0.0571
0.0274
0.000
0.000
0.000
0.000
(8)
Marginal
Effect
(=4)
0.014***
(0.005)
0.002***
(0.001)
-0.013*
(0.007)
-0.007
(0.009)
-0.004
(0.007)
0.002
(0.010)
-0.007
(0.016)
0.002
(0.004)
No
Yes
5
46,795
Robust standard errors are presented in parentheses. They are clustered by country and corresponding survey wave. These
are the results of the second stage regressions. In the first stages, education is explained by all the other controls presented
here and the excluded instruments: family savings and number of children. The results of the first stages are presented in
Appendix 4. Marginal effects measure the change in the probability of choosing the fourth category, for columns (1) to (6),
and the 10th category, for columns (7) and (8). The row called “Waves” specifies the survey waves included in the sample.
In column (1), for example, the regression only includes the 3rd and 4th waves of the survey. See appendix 1 for
correspondence of survey wave and year of survey. *** p<0.01, ** p<0.05, * p<0.1
30
Appendix 1. Country Sample by Survey Wave
Country Albania Algeria Argentina Armenia Australia Austria Azerbaijan Bangladesh Belarus Belgium Brazil Bulgaria Burkina Faso Canada Chile China Colombia Croatia Cyprus Czech Republic Denmark Dominican Republic Egypt, Arab Rep. El Salvador Estonia Ethiopia Finland France Germany Ghana Greece Hong Kong, China Hungary Iceland India Indonesia Iran, Islamic Rep. Iraq Ireland Israel Italy Japan Jordan Wave Total Country Wave 3 1994‐
1999 999 0 1,079 2,000 2,048 0 2,002 1,525 2,092 0 1,149 1,072 0 0 1,000 2,280 6,025 1,196 0 1,147 Wave 4 1999‐
2004 1,000 1,282 1,280 0 0 1,522 0 1,500 1,000 1,912 0 1,000 0 1,931 1,200 1,000 0 1,003 0 1,908 Wave 5 2005‐
2008 0 0 1,002 0 1,421 0 0 0 0 0 1,500 1,001 1,534 0 1,000 3,242 3,025 0 1,050 0 1,999 1,282 3,361 2,000 3,469 1,522 2,002 3,025 3,092 1,912 2,649 3,073 1,534 1,931 3,200 6,522 9,050 2,199 1,050 3,055 Korea, Rep. Latvia Lithuania Luxembourg Malaysia Mali Malta Mexico Moldova Morocco Netherlands New Zealand Nigeria Norway Pakistan Peru Philippines Poland Portugal Romania 0 1,023 0 1,023 417 0 0 417 Russian Federation Saudi Arabia 0 3,000 3,051 6,051 Singapore 1,254 1,021 0 987 0 2,026 0 0 0 0 1,005 0 1,038 1,615 2,036 0 1,142 0 0 0 1,500 1,014 1,001 2,064 1,534 0 1,252 1,254 2,026 1,500 3,039 2,616 6,126 1,534 1,142 1,252 650 0 2,040 0 0 1,000 968 2,002 1,004 2,532 0 0 2,001 2,015 2,667 0 0 0 0 1,054 0 2,325 1,012 1,199 2,000 1,362 1,223 2,701 0 0 1,012 1,096 1,200 Wave Total Wave 3 1994‐
1999 1,249 1,200 1,009 0 0 0 0 2,364 984 0 0 1,201 1,996 1,127 733 1,211 1,200 1,153 0 1,239 Wave 4 1999‐
2004 1,200 1,013 1,018 1,211 0 0 1,002 1,535 1,008 2,264 1,003 0 2,022 0 2,000 1,501 1,200 1,095 1,000 1,146 Wave 5 2005‐
2008 1,200 0 0 0 1,201 1,534 0 1,560 1,046 1,200 1,050 954 0 0 0 1,500 0 1,000 0 1,776 3,649 2,213 2,027 1,211 1,201 1,534 1,002 5,459 3,038 3,464 2,053 2,155 4,018 1,127 2,733 4,212 2,400 3,248 1,000 4,161 2,040 2,500 2,033 6,573 0 1,502 0 1,502 0 1,512 0 1,512 1,095 1,007 2,935 1,211 1,009 1,212 0 0 0 1,331 1,006 3,000 2,409 1,015 0 1,171 0 0 0 1,037 2,988 1,200 1,003 1,241 0 1,534 1,002 2,426 3,050 8,923 4,820 3,027 2,453 1,171 1,534 1,002 1,650 968 6,043 3,019 5,199 Slovak Republic Slovenia South Africa Spain Sweden Switzerland Tanzania Thailand Trinidad and Tobago Turkey Uganda Ukraine United Kingdom United States 1,907 0 2,811 1,093 1,542 4,607 1,002 1,195 2,000 1,200 1,346 0 1,000 1,041 1,249 7,860 1,002 5,006 4,134 3,991 5,026 1,012 1,199 3,012 3,512 2,423 Uruguay Venezuela, RB Vietnam Zambia Zimbabwe Total 1,000 1,200 0 0 0 71,791 0 1,200 1,000 0 1,002 94,894 0 0 1,495 1,500 0 72,573 1,000 2,400 2,495 1,500 1,002 239,258
31
Appendix 2. Education and Democracy for the Least Democratic Countries
Ordered Probit Regressions
Dependent variable: Approval of a democratic political system
ICRG - Democratic Accountability <= 3
(1)
(2)
Coefficient
Marginal Effect (=4)
Schooling
0.011***
0.004***
(0.004)
(0.002)
Age
0.002*
0.001*
(0.001)
(0.000)
Gender: Female
-0.037
-0.015
(0.023)
(0.009)
Marital Status: Single
0.006
0.002
(0.024)
(0.009)
Living with her parents
0.004
0.002
(0.029)
(0.012)
Employment status:
0.024
0.009
Employed
(0.031)
(0.012)
Employment status:
-0.010
-0.004
Unemployed
(0.041)
(0.016)
Scale of income
0.019***
0.007***
(0.007)
(0.003)
Country dummies
Yes
Yes
Year dummies
Yes
Yes
Observations
25,599
25,599
This sub-sample includes those countries with the lower “Democratic Accountability Index” (from
ICRG): Saudi Arabia, Iraq, VietNam, Pakistan, Belarus, Zimbabwe, Algeria, Egypt, Singapore, Uganda,
China, Nigeria, Hong Kong, and the Russian Federation. Definitions of the variables are shown in Table
1. Marginal effects measure the change in the probability of choosing the fourth category. Robust
standard errors are presented in parentheses. The row called “Waves” specifies the survey waves included
in the sample. See appendix 1 for correspondence of survey wave and year of survey. We replicated this
same exercise at the country level and find similar results for several countries, although not for all,
among the latter, China, Pakistan, and Saudi Arabia. Results are clustered by country and wave.
*** p<0.01, ** p<0.05, * p<0.1.
32
Appendix 3. Measures of Preference for Democracy and Education by Sub-Samples
Schooling
Age
Gender: Female
Marital Status: Single
Living with her parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
Country dummies
Year dummies
Waves
Observations
Country Waves
Pseudo R2
Ordered Probit Regressions
Dependent variable: Approval of a democratic political system
Satisfaction with democracy is <=
Satisfaction with
Democracy in the country is
Democracy in the country is >
mean (2.4)
democracy is > mean (2.4)
<= mean (6.4)
mean (6.4)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Coefficient
Marginal Effect Coefficient
Marginal
Coefficient
Marginal
Coefficient
Marginal Effect
(=4)
Effect (=4)
Effect (=4)
(=4)
0.036***
0.014***
0.033***
0.013***
0.025***
0.010***
0.026***
0.010***
(0.005)
(0.002)
(0.004)
(0.002)
(0.004)
(0.002)
(0.004)
(0.002)
0.003***
0.001***
0.004***
0.002***
0.005***
0.002***
0.004***
0.001***
(0.001)
(0.000)
(0.001)
(0.000)
(0.001)
(0.000)
(0.001)
(0.000)
-0.073***
-0.028***
-0.092***
-0.036***
-0.055***
-0.022***
-0.028
-0.011
(0.016)
(0.006)
(0.022)
(0.009)
(0.019)
(0.007)
(0.017)
(0.007)
0.004
0.002
0.026
0.010
0.024
0.010
-0.008
-0.003
(0.021)
(0.008)
(0.025)
(0.010)
(0.027)
(0.011)
(0.029)
(0.011)
-0.004
-0.002
-0.014
-0.006
0.013
0.005
-0.004
-0.002
(0.017)
(0.007)
(0.032)
(0.013)
(0.029)
(0.011)
(0.028)
(0.011)
-0.018
-0.007
-0.054**
-0.021**
-0.016
-0.006
-0.029
-0.011
(0.019)
(0.007)
(0.022)
(0.009)
(0.023)
(0.009)
(0.028)
(0.011)
-0.077***
-0.029***
-0.095**
-0.037**
0.000
0.000
-0.012
-0.005
(0.027)
(0.010)
(0.037)
(0.015)
(0.035)
(0.014)
(0.038)
(0.015)
0.019***
0.007***
0.013**
0.005**
0.006
0.002
0.004
0.001
(0.004)
(0.002)
(0.006)
(0.003)
(0.008)
(0.003)
(0.008)
(0.003)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
3, 4
3, 4
3, 4
3, 4
5
5
5
5
37,921
37,921
33,898
33,898
21,321
21,321
24,326
24,326
67
67
37
37
0.0631
0.0574
0.0554
0.0588
To estimate the sub-samples, we generate the media of “satisfaction with democracy” and “democracy in the country”, and they were considered
the cut-off points. Their values are show in the Table. Marginal effects measure the change in the probability of choosing the fourth category.
Robust standard errors are presented in parentheses. They are clustered by country and wave. The row called “Waves” specifies the survey waves
included in the sample. See Appendix 1 for correspondence of survey wave and year of survey. *** p<0.01, ** p<0.05, * p<0.1.
33
Appendix 4
Measures of Preference for Democracy and Education (Marginal Effects)
Schooling
Age
Gender: Female
Marital Status:
Single
Living with her
parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
Size of town
Respect of human
rights in own
country
Self-positioning in
political scale
Satisfaction with the
way democracy
develops
Democracy in own
country
Agreem. with "In
democracy, economic
system does not run
badly" (=4)
Ordered Probit Regressions (Marginal Effects)
Dependent variable:
Agreem. with
Agreem. with
"Democracies are
"Democracy may have
good at maintining
problems but is better"
order" (=4)
(=4)
Importance of living
in a country that is
governed
democratically (=10)
(1)
0.006***
(0.001)
0.000
(0.000)
-0.018***
(0.002)
0.009**
(0.003)
-0.009***
(0.003)
0.002
(0.002)
-0.015***
(0.003)
0.007***
(0.001)
(3)
0.006***
(0.001)
-0.000***
(0.000)
-0.010***
(0.002)
0.004
(0.003)
-0.009***
(0.003)
0.004
(0.002)
-0.008**
(0.003)
0.006***
(0.001)
(7)
0.012***
(0.001)
0.002***
(0.000)
-0.012*
(0.006)
-0.001
(0.007)
0.000
(0.006)
0.001
(0.007)
-0.009
(0.014)
0.003
(0.003)
(2)
0.007***
(0.001)
0.000
(0.000)
-0.019***
(0.003)
0.007*
(0.004)
-0.011***
(0.004)
0.001
(0.004)
-0.015***
(0.006)
0.007***
(0.001)
0.001
(0.001)
0.018***
(0.004)
0.003*
(0.002)
0.027***
(0.005)
(4)
0.007***
(0.001)
-0.000***
(0.000)
-0.007**
(0.003)
-0.000
(0.004)
-0.010**
(0.004)
0.004
(0.003)
-0.010*
(0.006)
0.005***
(0.001)
0.002**
(0.001)
0.019***
(0.004)
-0.001
(0.002)
0.027***
(0.004)
(5)
0.010***
(0.001)
0.002***
(0.000)
-0.018***
(0.004)
-0.005
(0.006)
0.003
(0.005)
0.001
(0.005)
-0.011
(0.008)
0.007***
(0.002)
(6)
0.012***
(0.002)
0.002***
(0.000)
-0.019***
(0.005)
-0.013
(0.009)
0.010
(0.008)
-0.007
(0.007)
-0.018
(0.012)
0.007***
(0.002)
0.005***
(0.002)
0.039***
(0.009)
0.008**
(0.004)
0.060***
(0.010)
(8)
0.016***
(0.002)
0.003***
(0.001)
-0.013
(0.010)
-0.008
(0.012)
-0.001
(.)
-0.006
(0.011)
-0.028
(0.022)
-0.001
(0.004)
0.005
(0.003)
0.011
(0.012)
0.007
(0.005)
0.039***
(0.006)
Waves
3,4
3,4
3,4
3,4
3,4
3,4
5
5
Observations
105,417
42,415
107,692
42,914
108,941
43,520
52,766
23,310
Marginal effects measure the change in the probability of choosing the fourth category, for columns (1) to (6), and the 10th
category, for columns (7) and (8). Robust standard errors are presented in parentheses. They are clustered by country and
wave. The row called “Waves” specifies the survey waves included in the sample. See appendix 1 for correspondence of
survey wave and year of survey. *** p<0.01, ** p<0.05, * p<0.1.
34
Appendix 5. First Stages of Tables 8 and 9
Age
Gender: Female
Marital Status: Single
Living with her
parents
Employment status:
Employed
Employment status:
Unemployed
Scale of income
Number of children
Savings: Family just
got by
Savings: Family spent
some savings
Savings: Family spent
savings and borrowed
money
Constant
OLS regressions
Dependent variable: Years of schooling
Table 7
Table (8)
(1)
(3)
(1)
(3)
(5)
-0.038***
-0.041***
-0.035***
-0.035***
-0.036***
(0.004)
(0.007)
(0.005)
(0.005)
(0.005)
-0.144**
-0.044
-0.044
-0.061
-0.076
(0.058)
(0.083)
(0.065)
(0.065)
(0.067)
0.111
0.119
0.006
-0.014
0.013
(0.084)
(0.130)
(0.120)
(0.116)
(0.118)
-0.015
-0.008
-0.040
-0.015
-0.023
(0.056)
(0.074)
(0.078)
(0.077)
(0.074)
0.861***
0.792***
0.850***
0.817***
0.828***
(0.095)
(0.117)
(0.126)
(0.125)
(0.133)
0.143
-0.031
0.243*
0.204
0.229*
(0.111)
(0.166)
(0.128)
(0.124)
(0.134)
0.459***
0.350***
0.445***
0.453***
0.454***
(0.029)
(0.033)
(0.036)
(0.036)
(0.037)
-0.456***
-0.389***
-0.410***
-0.420***
-0.414***
(0.029)
(0.040)
(0.033)
(0.033)
(0.034)
-0.646***
-0.787***
-0.531***
-0.519***
-0.534***
(0.056)
(0.088)
(0.063)
(0.067)
(0.068)
-0.352***
-0.576***
-0.321***
-0.349***
-0.323***
(0.087)
(0.127)
(0.103)
(0.105)
(0.106)
-0.659***
-0.776***
-0.622***
-0.587***
-0.628***
(7)
-0.044***
(0.006)
-0.256**
(0.114)
0.093
(0.132)
0.082
(0.087)
0.820***
(0.122)
0.064
(0.212)
0.461***
(0.039)
-0.514***
(0.052)
-0.748***
(0.099)
-0.359**
(0.160)
-0.752***
(0.091)
(0.131)
(0.123)
(0.120)
(0.127)
(0.169)
9.078***
8.094***
11.766***
11.551***
11.778***
12.901***
(1.026)
(0.401)
(0.606)
(0.634)
(0.632)
(0.412)
Observations
127,011
39,090
72,215
73,745
74,076
46,566
R-squared
0.329
0.330
0.305
0.307
0.317
0.365
Robust standard errors are presented in parentheses. They are clustered by country and corresponding survey wave.
These are the results of the first stage regressions of Tables 7 and 8 (using IV). Education is explained by all the
other controls presented here and the excluded instruments: family savings and number of children.
*** p<0.01, ** p<0.05, * p<0.1
35